Look into alternative "FastStabbing" implementation.

[joshuak94]

FastStabbing is an implementation for interval range queries which assumes n intervals with left and right endpoints in the range [1,...,O(n)]. Then it can index this list in Ω(n) space and time, with queries solved in Ω(1 + k) time, where k = # of results produced by the query.

This is better than Interval Trees, which require Ω(n log n) time to build and Ω(log n + k) time to query.